|Publication number||US7580564 B2|
|Application number||US 10/844,787|
|Publication date||Aug 25, 2009|
|Filing date||May 13, 2004|
|Priority date||May 13, 2004|
|Also published as||US20050254706|
|Publication number||10844787, 844787, US 7580564 B2, US 7580564B2, US-B2-7580564, US7580564 B2, US7580564B2|
|Inventors||James Ray Bailey, Curt Paul Breswick, David Allen Crutchfield, Thomas Jon Eade, Zachary Nathan Fister|
|Original Assignee||Lexmark International, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (18), Referenced by (1), Classifications (8), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The invention relates to a method of transforming data, and particularly, to a method of transforming data during a tonal transformation process.
An image scanned using an imaging device (e.g. a scanner, a multi-functional device such as a scanner-printer-facsimile machine, etc.) may have several transformations performed on it before the resultant image is either displayed (e.g., on a PC) or printed (e.g., via a printer). For example, the target image can be scanned at a bit-depth of 30 to 48 bits, and then transformed to a 24-bit image where the 24-bit image is saved and viewed on a personal computer. Alternately, the 24-bit image can be processed to a 4-bit to 9-bit image and printed on a printing device.
One common type of image transformation is known as tonal transformation. A scanned image may go through several tonal transformations to achieve the desired result. Example tonal transformations include gamma compensation, brightness/contrast adjustment, and shadow enhancement. The quality of the transformation has a direct effect on the quality of the scanned output image, making tonal transformation important to the quality of a scanner. Poor methods of transformation may cause visually detectable quantization in the scanned output image and a decreased scanner modulated transfer function (MTF) measurement. One method to reduce quantization is to input a higher bit-depth image than is output during the tonal correction (e.g. a 36-bit image becomes a 24-bit image after the tonal transformation).
One method of tonal transformation is to apply a mathematical equation to each pixel in the image. This is a common way to transform a RGB image to a sRGB image or to adjust the brightness or contrast of an image. In multi-functional devices, a transformation such as this usually takes place inside the application specific integrated circuit (ASIC) during a standalone copy. Applying multiple, complex mathematical equations to the image for such transformations may be detrimental to the performance of the copy operation and may be too inflexible to be practically implemented in an ASIC.
To maintain flexibility and performance of the ASIC, a lookup table (LUT) can be used to perform tonal transformations. In one method, the value of the pixel indexes directly into a table with the resulting output pixel being returned. What is stored in the table can be easily updated, thus allowing for flexibility. Multiple transformations can be achieved using a single LUT, and thus, further improve performance.
One problem with using a direct indexing method is the size of the LUT. For example, if a 16-bit color indexes into a table that returns an 8-bit color, a 65536×8 table is required. This would result in a 64 KB table. For a 48-bit to 24-bit tonal table (or three 16-bit to 8-bit tables, each corresponding to one of three colors (e.g., red, green, and blue)), the table requires a 192 KB RAM, which is costly to implement in hardware. Placing the table in main memory and using direct memory access (DMA) to access the table places a significant burden on memory resources, which may affect the performance of the ASIC. It would be beneficial to have an alternative method of performing tonal transformation.
In one embodiment, the invention provides a method of transforming a n-bit data packet to a m-bit data packet with a lookup (LUT) table. The LUT table includes at least one entry data packet and at least one respective delta value associated with each entry data packet. The method includes the acts of receiving an input data packet having n-bits, indexing the LUT table with at least a portion of the input data packet to obtain one of the at least one entry data packet, and decompressing the obtained entry data packet with the at least one respective delta value associated with the obtained entry data packet, thereby resulting in an output data packet having m-bits. The decompressing act includes using a portion of the input data packet to determine the number of delta values called for decompressing the obtained entry data packet. The method can be used in, for example, an image processor.
In another embodiment, the invention provides an image processor for transforming a n-bit image data packet to a m-bit data packet. The image processor includes a memory having one or more LUT tables. The one or more LUT tables include z entry values and c=(2n)/z−1 respective delta values associated with each entry value. The c respective delta values are represented by Δ(1) . . . Δ(c). The image processor includes a processor configured to transform a binary input value represented by input value (n−1:0) to a binary output value represented by output value (m−1:0). The processor indexes the LUT table with the bits input value (n−1:n−k) to obtain an associated entry value represented by entry value [x], where k is equal to k=ceiling(log2(z)). The processor includes a decompressor configured to decompress the entry value [x] to obtain the output value (m−1:0). The decompression uses the equation
where y represents base10 [input value (n−k−1:0)].
Other features and aspects of the invention will become apparent by consideration of the detailed description and accompanying drawings.
Before any embodiments of the invention are explained in detail, it is to be understood that the invention is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the following drawings. The invention is capable of other embodiments and of being practiced or of being carried out in various ways. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The use of “including,” “comprising,” or “having” and variations thereof herein is meant to encompass the items listed thereafter and equivalents thereof as well as additional items. Unless specified or limited otherwise, the terms “mounted,” “connected,” “supported,” and “coupled” and variations thereof are used broadly and encompass both direct and indirect mountings, connections, supports, and couplings. Further, “connected” and “coupled” are not restricted to physical or mechanical connections or couplings.
With reference to
Each line sensor 125, 130, and 135 charges to a voltage level corresponding to the intensity of the color detected for that element. The voltage for each element of the captured line is then shifted out of the image sensor serially and sent to an analog front-end device 140 (
Before proceeding further, it should be noted that the scanner 100 includes other components not shown or described herein. For example, the scanner 100 includes a scanner motor to move the light source 105, optics 115, and sensor 120 across the target image. It should also be noted that the scanner discussed in
Typically, the scanner 100 performs the data transformation during a tonal transformation process, examples of which are discussed below. However, the invention can be used in other applications that require data transformation (or compression) and is not limited to tonal transformation processes.
Discussed below are three Example tonal transformation processes. Example 1 provides an example of a tonal transformation of the prior art. Examples 2 and 3 provide examples of tonal transformations incorporating the invention. Examples 1 and 2 are described in connection with a 36-bit to 24-bit tonal transformations. However, as will become apparent in Example 3, the invention is not limited to 36-bit to 24-bit tonal transformations and the invention can be applied to other data transformation applications.
Three 12-bit to 8-bit tonal transformation tables 150, 155, and 160 of the prior art are schematically represented in
For the construction shown, n is equal to twelve and m is equal to eight, resulting in 32,768 storage elements, or 4 KB worth of SRAM. Thus, a 12 KB SRAM is used to store the three tables 150, 155, and 160. The tables 150, 155, and 160 provide a flexible, high performance means for tonal transformation and are used to intelligently truncate 36-bit scan data to 24-bit scan data while minimizing quantization errors. However, the tables 150, 155, and 160 require approximately 0.5 mm2 worth of die area to implement in a typical 0.13 μm cmos process.
One type of tonal transformation used in scanning is gamma compensation. An example of a typical gamma curve 165 is shown in
In one embodiment of the invention, the above two characteristics of the tonal transformation data are used to compress the lookup table to a much smaller tonal transformation table. Specifically, a sparsely populated table (as compared to Example 1) and the associated delta values between entries are stored in one or more tables. An example of a table 170 used for performing a red element transformation is shown in
To use the table 170, the upper eight bits of the 12-bit input entry (represented by red (11:4)) are used to index into the table 170. What is returned is an 8-bit value plus fifteen delta entries (represented by compressed red entry (22:0)). The lower four bits of the 12-bit input pixel (represented by red (3:0)) are then used to select how many delta values are added to the returned 8-bit entry value. For example, if the input value is Red (100000000101), then the entry value at address 10000000 is returned from table 170 along with the associated fifteen delta values. Using the lowest four bits of the input value, 0101, it is determined that the first five delta values are added to the returned entry value. If the entry returned was 75base 10 and Δ1=0, Δ2=1, Δ3=0, Δ4=0, and Δ5=1, then the resulting pixel is 77base10. In general, the output is based on the equation:
where x=input pixel(11:4) and y=input pixel(3:0).
The discussion herein will focus on the red table 170. The ASIC receives an input value represented by input value (11:0). The first 8 bits input value (11:4) are used to index table 170. As a result of the application, the ASIC receives entry value [x] and the delta values Δ1 . . . Δ15. The compressed red entry (22:0) and input value (3:0) are then applied to the decompressor 185. Using equation [e1], the decompressor 185 modifies the entry value with the appropriate delta values, resulting in output value (7:0).
As discussed above, Example 2 implements a 265×23 table. However, the described transformation technique can be generically applied. Additionally, for Example 2, the delta values were limited to +0 or +1. However, if a delta value greater than +1 or less than +0 were expected, then the bit width for each delta value would increase accordingly. For example, if values of +1, 0, and −1 are expected, then 2-bit delta values can be used, where the first bit is a sign bit and the second bit is a value bit. Other variations are possible.
If n equals the number of input bits and m equals the number of output bits, then a fully populated table (e.g., as shown in Example 1) requires 2n*m storage bits. For a compressed tonal table using the technique described herein, the corresponding table will have the following number of storage bits:
where z equals the number of table entries, and b equals the number of bits in each delta value.
Referring to example 2, the number of table entries was 256, the number of bits in each delta value was 1, the number of input bits was 12 and the number of output bits was 8. Substituting into equation [e2]
The number of table entries defines how sparsely populated the table is. A least sparsely populated table would contain a single entry plus a series of delta values that correspond to the remaining entries. However, this would require an addition operation for each delta entry to decompress the table in hardware. In the implementation of Example 2, a 256-entry table was selected with fifteen delta values corresponding to the values between entries. This translates to a maximum of fifteen addition operations to decompress the corresponding tonal value from the table. A more sparsely populated table would result in a smaller table, and thus more addition operations would be required to compute the decompressed output value.
The input address into the compressed table is based on the number of table entries. The upper k bits of the input pixel will correspond to the input address into the table where
Referring to example 2, z was 256. Substituting into equation [e4],
The lower n−k bits of the input pixel are used to determine how many delta values are added to the returned entry. Referring to example 2, n−k equals 4 bits. This identifies that the lower four-bits of the input pixel are used to determine how many deltas to add to the entry value to obtain the output value for the corresponding input value.
The compression technique of Examples 2 and 3 is lossless as long as the programmed table follows the two assumptions below. The first assumption is the output always increases or stays the same as the input increases. That is, the output never decreases as the input increases. Second, when applying the transformation technique, the difference between neighboring output entries in the table differs by a known value. The result is a compression technique that reduces the data size required for a tonal transformation (e.g., an eighty-two percent reduction was shown from Example 1 to Example 2). This savings could be realized in the amount of die area consumed by a transformation table within the ASIC or it could be realized in the amount of disk space required by the PC to store all of the possible transformation tables required by a product.
While Examples 2 and 3 were described as obtaining the entry value and all associated delta values, it is envisioned that other variations are possible. For example, it is envisioned that the ASIC analyzes input value (n−k−1:0) prior to decompression and only obtains the appropriate delta values. More specifically and with reference to table 170, the ASIC can look at the first 8 bits to index the appropriate entry value and uses the last four bits to obtain the appropriate delta values. The decompressor can then decompress the resulting information. It is also envisioned that each lookup table can be divided across multiple tables.
Thus, the invention provides, among other things, a new and useful method of transforming a n-bit data packet to a m-bit data packet using a lookup table. Various features and advantages of the invention are set forth in the following claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5172237||Dec 19, 1991||Dec 15, 1992||Ricoh Corporation||High quality color image compression system|
|US5479587||Sep 3, 1992||Dec 26, 1995||Hewlett-Packard Company||Page printer having adaptive data compression for memory minimization|
|US5504591||Apr 25, 1994||Apr 2, 1996||Microsoft Corporation||System and method for compressing graphic images|
|US5517335||Oct 11, 1994||May 14, 1996||Seiko Epson Corporation||Apparatus and method for enhancing color saturation in halftoned images|
|US5596510||Mar 23, 1994||Jan 21, 1997||Island Graphics||Table-based linear interpolation for color correction system and method|
|US5623556||Jan 4, 1995||Apr 22, 1997||Kabushiki Kaisha Toshiba||System and method of extracting binary image data|
|US5687300||Mar 27, 1995||Nov 11, 1997||Linotype-Hell Ag||Process for creating five to seven color separations used on a multicolor press|
|US5724450||Sep 30, 1994||Mar 3, 1998||Apple Computer, Inc.||Method and system for color image compression in conjunction with color transformation techniques|
|US5768481||Jun 5, 1995||Jun 16, 1998||Apple Computer, Inc.||Method and apparatus for compression of digitized image data using a dynamic band recompression scheme|
|US5930387||Dec 4, 1997||Jul 27, 1999||Apple Computer, Inc.||Method and apparatus for encoding color image data using dynamic color matching|
|US5983251||Mar 8, 1996||Nov 9, 1999||Idt, Inc.||Method and apparatus for data analysis|
|US6393148||May 13, 1999||May 21, 2002||Hewlett-Packard Company||Contrast enhancement of an image using luminance and RGB statistical metrics|
|US6510247||Dec 17, 1998||Jan 21, 2003||Hewlett-Packard Company||Decoding of embedded bit streams produced by context-based ordering and coding of transform coeffiecient bit-planes|
|US6580825||Dec 22, 2001||Jun 17, 2003||Hewlett-Packard Company||Contrast enhancement of an image using luminance and RGB statistical metrics|
|US6665454||Jul 10, 1998||Dec 16, 2003||Silverbrook Research Pty Ltd||Dot adjacency compensation in optical storage systems using ink dots|
|US6774953 *||May 31, 2001||Aug 10, 2004||Sony Corporation||Method and apparatus for color warping|
|US6828982 *||Oct 31, 2002||Dec 7, 2004||Samsung Electronics Co., Ltd.||Apparatus and method for converting of pixels from YUV format to RGB format using color look-up tables|
|US20030016305 *||May 31, 2001||Jan 23, 2003||Mark Champion||Method and apparatus for color warping|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US20080269481 *||Feb 4, 2008||Oct 30, 2008||Laval Chan Chun Kong||Compounds and methods for the treatment or prevention of Flavivirus infections|
|U.S. Classification||382/166, 382/251|
|International Classification||H04N1/60, G06K9/46, G06K9/00, G06K9/36|
|May 13, 2004||AS||Assignment|
Owner name: LEXMARK INTERNATIONAL, INC., KENTUCKY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BAILEY, JAMES RAY;BRESWICK, CURT PAUL;CRUTCHFIELD, DAVIDALLEN;AND OTHERS;REEL/FRAME:015331/0944
Effective date: 20040505
|Jan 30, 2013||FPAY||Fee payment|
Year of fee payment: 4